CN106845866A - Equipment method for predicting residual useful life based on improved particle filter algorithm - Google Patents
Equipment method for predicting residual useful life based on improved particle filter algorithm Download PDFInfo
- Publication number
- CN106845866A CN106845866A CN201710107434.5A CN201710107434A CN106845866A CN 106845866 A CN106845866 A CN 106845866A CN 201710107434 A CN201710107434 A CN 201710107434A CN 106845866 A CN106845866 A CN 106845866A
- Authority
- CN
- China
- Prior art keywords
- particle
- current time
- variance
- filter algorithm
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to the failure predication field of electromechanical equipment, a kind of equipment method for predicting residual useful life based on improved particle filter algorithm is disclosed, improve the precision of equipment life prediction.Improved particle filter algorithm of the present invention includes importance sampling stage and resampling stage, and Unscented kalman filtering method more new particle, generation suggestion distribution, so as to suppress sample degeneracy are introduced in the importance sampling stage；Increase Markov chain Monte-Carlo step in the resampling stage, suppress sample exhausted.The present invention is applied to lithium battery, rolling bearing, gear-box.
Description
Technical field
It is surplus the present invention relates to the failure predication field of electromechanical equipment, more particularly to the equipment based on improved particle filter algorithm
Remaining life-span prediction method.
Background technology
With human civilization progress and scientific and technological accelerated development, aircraft, ship, power generating equipment, Transit Equipment etc.
The safety of great dynamoelectric equipment is on active service and be significant for Chinese national economy development and national defense safety.Yet with reality
Operating mode is complicated and changeable, bad environments, and equipment can be gradually aging in long-term running, and residual life can be gradually reduced, easily
Serious accident is produced, great property loss or casualties is caused；And carry out that if maintenance is changed surplus can be repaired if blindness, make
Into huge waste.Therefore, accurate life prediction, for ensureing equipment safety, Effec-tive Function, increases economic efficiency with weight
Want meaning.
In recent years, occur in that many methods that can be used for the prediction of electromechanical equipment remaining life, such as autoregression (AR,
Auto Regressive) temporal model method, artificial neural network (ANN, Artificial Neural Networks), branch
Hold vector machine (SVM, Support Vector Machine), Method Using Relevance Vector Machine (RVM, Relevance Vector Machine)
Deng based on data-driven Forecasting Methodology, and directly to the specific parts Analytic modeling of product or emulation be predicted based on
Physical Modeling etc..
Wherein, particle filter (PF, Particle Filter) is a kind of pre- with the equipment life of data mixing based on model
Survey method, compared to traditional data-driven Forecasting Methodology, the method can provide the uncertain expression for predicting the outcome, this for
It is particularly significant for policymaker.In addition, its model being based on is different from general physical model, but a kind of experience degeneration mould
Type.The model seeks the gathered data rule for changing over time or the recursion for seeking internal system state from data Angle
Relation, this Mathematical Modeling is also wider compared to acquisition, the scope of application is easier.Meanwhile, incorporated experience into using particle filter method
Degradation model realizes predicting residual useful life, has good adaptability for non-gaussian, non-linear degradation process.At present, base
In the life-span prediction method of particle filter successful Application has been obtained in lithium battery, rolling bearing, gear-box etc..
Even so, such is based on the life-span prediction method of particle filter algorithm, two main defects are presently, there are, i.e.,
Sample degeneracy is exhausted with sample, causes equipment predicting residual useful life precision not high.
The content of the invention
The technical problem to be solved in the present invention is：There is provided a kind of equipment residual life based on improved particle filter algorithm pre-
Survey method, improves the precision of equipment life prediction.
To solve the above problems, the technical solution adopted by the present invention is：Equipment based on improved particle filter algorithm is remaining
Life-span prediction method, the particle filter algorithm for being used includes importance sampling stage and resampling stage, it is characterised in that
The importance sampling stage introduces Unscented kalman filtering (UKF, Unscented Kalman Filter) method more new particle, raw
It is distributed into suggestion, so as to suppress sample degeneracy；Increase Markov chain Monte-Carlo (MCMC, Markov in the resampling stage
Chain Monte Carlo) step, suppress sample exhausted.
Further, the step of generation suggestion distribution includes：
The initial variance of state value and particle according to previous moment particle calculates the sigma point sets of previous moment particle；
Bring the point that the sigma points of previous moment particle are concentrated into particle state equation, obtain the particle at current time
Sigma point sets；
Particle sigma point set weighted sums to current time, so as to obtain the expectation of current time particle；
The expectation of current time particle and the variance of the sigma point sets of current time particle are done, and gained variance is carried out
Weighted sum, obtains the time updated value of the variance of current time particle；
The point that the point and the sigma points of previous moment particle that the sigma points of current time particle are concentrated are concentrated is brought into
Particle state equation, so as to obtain the observation of current time particle；
Observed value weighting summation to the particle at current time, obtains the updated value of the observation of current time particle；
Particle sigma point sets based on current time, the expectation of current time particle, the observation of the particle at current time
Value, the updated value of the observation of current time particle, are calculated the measurement updated value of the variance of the particle at current time；
Calculate the correction factor of the variance of current time particle；
The time updated value of the variance based on current time particle, the measurement updated value of the variance of the particle at current time,
The correction factor of the variance of current time particle, is calculated the correction value of the variance of current time particle；
With the expectation of current time particle, the expectation and side of the correction value as Gaussian Profile of the variance of current time particle
Difference, generation suggestion point.
Further, Markov chain Monte-Carlo step includes：
Threshold value u~U [0,1] is extracted, sampling obtains the candidate state of particle from suggestion distributionUsing the time of particle
Select stateIts acceptance probability A is calculated, if u≤A, is receivedI.e.：Otherwise abandonCarry out resampling and protect
Stay the particle of resamplingI.e.：
Specifically, entire protocol of the invention includes：
(1) forecast model is set up；
(2) degraded performance is assessed, and chooses sensitive features as degenerative character；
(3) the original degenerative character chosen is pre-processed；
(4) on the basis of forecast model, prior information is obtained from the historical data of degenerative character；
(5) according to the initial information of each parameter in degradation model distribution and model, improved particle filter algorithm is carried out just
Beginningization；
(6) estimation and two processes of outside forecast are iterated to each moment successively using improved particle filter algorithm,
When degenerate state predicted value is less than or equal to failure threshold, prediction terminates；
(7) state to all particle samples is counted, and obtains final predicting the outcome.
Further, in the forecast model of step (1), the system mode renewal equation and measurement equation for predicting object are distinguished
For：
xk=fk(xk-1,vk-1)
zk=hk(xk,nk)
Wherein, xkRepresent the system mode value at current time, zkExpression current time contains the measuring value of additive noise, vk-1
It is the process noise of previous moment, nkIt is that the measurement noise at current time, k ∈ N, N particle numbers, and process noise are made an uproar with measurement
Sound is stationary noise sequence.
Further, step (3) carries out the pretreatment of smooth and dullnessization to primitive character.
Further, prior information include degenerative character with the distribution map and forecast model of variable each parameter it is initial
Value.
The beneficial effects of the invention are as follows：The present invention introduces UKF methods by the importance sampling stage, and standard particle is filtered
The importance density function of ripple algorithm is improved, and suppresses sample degeneracy, and MCMC steps are increased in the resampling stage, suppresses sample withered
Exhaust, compared to standard particle filter forecasting method, precision of prediction is higher, can especially provide more accurate early prediction.
Brief description of the drawings
Fig. 1 is institute's improved particle filter algorithm flow chart of the present invention；
Fig. 2 is that prior information obtains flow chart；
Fig. 3 is monitoring objective object lifetime prediction flow chart proposed by the invention；
Fig. 4-7 is respectively battery A1-A4 degradation in capacity True Data and curve matching；
Particle filter is improved when Fig. 8 is 18 groups of measurement data before known A4 to predict the outcome；
Particle filter is improved when Fig. 9 is 30 groups of measurement data before known A4 to predict the outcome
Standard particle filter forecasting compares knot with particle filter prediction is improved when Figure 10 is 18 groups of measurement data before known A4
Really；
Standard particle filter forecasting compares knot with particle filter prediction is improved when Figure 11 is 30 groups of measurement data before known A4
Really.
Numbered in figure：Un is failure threshold, and Tn iteration update cycles, PDF is the probability density function of embodiment, and PDF ＇ are
The probability density function of standard particle filtering.
Specific embodiment
Below from two parts --- innovatory algorithm part and equipment predicting residual useful life part, the present invention is carried out in detail
Illustrate.(1) improved particle filter algorithm
The improved particle filter algorithm introduces UKF methods by the importance sampling stage, to standard particle filtering algorithm
Importance density function be improved, suppress sample degeneracy, increase MCMC steps in the resampling stage, suppress sample exhausted.Calculate
The flow chart of method is as shown in figure 1, it is concretely comprised the following steps：
(1) particle initialization
For moment k=0, i=1,2 ..., N, j=1,2 ..., 2l, from prior probability p (x0) in randomly select N number of grain
Son is obtainedAnd make particle weights
Calculate average：
It is augmented treatment：
Calculate variance：
It is augmented and obtains：
(2) importance sampling
(1), for k=1,2 ..., moment utilize UKF more new particles, generation suggestion distribution.Comprise the following steps that：
1. particle sigma point sets are calculated.That is, according to the state value of previous moment particleAnd the initial variance of particle
Calculate the sigma point sets of previous moment particleFormula is as follows：
In formula, λ=α2(n+ κ)-n is a proportionality coefficient；The size of α determines the sample of selection around particle average's
Distribution situation, generally takes a small positive number；κ is another proportionality coefficient, can be taken as 0；Parameter beta reflects the priori on particle x
Distribution, for Gaussian Profile, β=2；Subduplicate i-th row of representing matrix.
2. time renewal：
A. the point sigma points of previous moment particle concentratedBring particle state equation into, obtain current time
Particle sigma point setsFormula is as follows：
B. to the particle sigma point sets at current timeWeighted sum, so as to obtain the expectation of current time particleFormula is as follows：
C. the expectation of current time particle is doneWith the sigma point sets of current time particleVariance, and to institute
Obtain variance and be weighted summation, obtain the time updated value of the variance of current time particleFormula is as follows：
D. the point sigma points of current time particle concentratedAnd the sigma points concentration of previous moment particle
PointParticle state equation is brought into, so as to obtain the observation of current time particleFormula is as follows：
E. to the observation of the particle at current timeWeighted sum, obtains the renewal of the observation of current time particle
ValueFormula is as follows：
3. measure and update (incorporating newest observation)：
I. the particle sigma point sets at current time are based onThe expectation of current time particleThe grain at current time
The observation of sonThe updated value of the observation of current time particleIt is calculated the variance of the particle at current time
Measurement updated valueFormula is as follows：
Ii. the adjusted coefficient K of the variance of current time particle is calculatedk, formula is as follows：
Iii. it is based on the time updated value of the variance of current time particleThe measurement of the variance of the particle at current time
Updated valueThe adjusted coefficient K of the variance of current time particlek, it is calculated the correction value of the variance of current time particleFormula is as follows：
4. with the expectation of current time particleThe correction value of the variance of current time particleAs the phase of Gaussian Profile
Hope and variance, ultimately generate suggestion distribution
(2) sample：Sampled from suggestion distribution
Importance weight is updated to：
Normalizing right value update is：
(3) resampling：
MCMC steps, extract threshold value u~U [0,1], are distributed from suggestionMiddle sampling obtains candidate stateMeter
Calculate acceptance probabilityIf u≤A, receiveI.e.：Otherwise abandon
Resampling is carried out according to formula (16) and retain the particle of resamplingI.e.：
(4) output result：
By the iteration of algorithm above, then the system state estimation of moment k is respectively with variance evaluation：
If k≤T0(T0It is the number of known measurements), then make k=k+1 and return to step (2) carries out importance weight more
Newly, down-stream is continued executing with.Otherwise, program, output result are terminated.
(2) equipment life prediction
Equipment predicting residual useful life is carried out using above-mentioned improved particle filter algorithm, is a kind of longevity based on model and data
Life Forecasting Methodology, its prediction process is as follows：
(1) forecast model is set up
Before application enhancements particle filter algorithm, first have to set up the degradation model of prediction object.Typically can basis
Known system physical knowledge sets up model, and model is set up also dependent on experience.It is mould to set up the model most important
Type must be healthy and strong, be capable of the change of accurate quick response system.In general predictive model, the system mode of object is predicted
Renewal equation is respectively with measurement equation：
xk=fk(xk-1,vk-1) (19)
zk=hk(xk,nk) (20)
Wherein, xkRepresent the system mode value at current time, zkExpression current time contains the measuring value of additive noise, vk-1
It is the process noise of previous moment, nkIt is that the measurement noise at current time, k ∈ N, N particle numbers, and process noise are made an uproar with measurement
Sound is stationary noise sequence.
(2) degraded performance assessment is chosen with sensitive features
" quality " of degenerative character can directly affect the order of accuarcy for predicting the outcome.So-called " quality " is referred to for predicting
Degenerative character reflect goal systems degradation trend ability.Generally performance degradation should be carried out according to equipment running status information
Assessment, selection can preferably reflect that the sensitive features that equipment performance is degenerated are used for predicting residual useful life.
(3) original sensitive features pretreatment
In actual prediction, the trend that degenerative character reflects system degradation is generally only concerned, without concern for degenerative character institute
Comprising other details information.Due to the influence of regulation outside noise etc., for original degenerative character, its degeneration scope is often difficult
To determine.In addition, the degenerative process of equipment is dull irreversible.So, the preliminary degenerative character extracted is carried out it is smooth and
The pretreatment of dullnessization, preferably can predict suitable for particle filter.
(4) prior information of historical data is obtained
On the basis of above-mentioned forecast model, both sides priori letter can be obtained from the historical data of degenerative character
Breath.One is situation of change of the degenerative character with variable (time or cycle etc.), for improving the general predictive model set up.Two are
The initial value of each parameter in forecast model, it is relevant with the initialization for improving particle filter forecast sample.The acquisition stream of prior information
Journey is as shown in Figure 2:
(5) prediction of monitoring object
After the prior information needed for acquisition, just can be with improved particle filter algorithm to the target pair of existing Monitoring Data
As being predicted.First, according to the initial information of each parameter in degradation model distribution and model, algorithm is initialized.So
Afterwards, estimation and two processes of outside forecast are iterated successively, when degenerate state predicted value is less than or equal to failure threshold, in advance
Survey terminates.Finally, the state to all particle samples is counted, and obtains final predicting the outcome.Its pre- flow gauge such as Fig. 3 institutes
Show.
Embodiment
Below in conjunction with instantiation --- lithium-ion battery systems state estimation and life prediction illustrate tool of the invention
Body implementation method.Because lithium ion battery operation principle is complicated, its internal state parameter is difficult to measure, typically by gathering battery
The data such as electric current, voltage in charge and discharge process, set up the Mathematical Modeling between capacity and cycle-index to predict its residual life.
Its detailed process is as follows：
(1) general predictive model is set up
Target prediction model is set up, the system mode renewal equation and measurement equation for predicting object are respectively formula (19),
(20) shown in.
(2) sensitive features are chosen
The same with most of secondary cells, during recycling, its capacity can be with cycle-index for lithium ion battery
Increase and it is less and less, once dropping to default critical value, battery will be regarded as failure, it is impossible to is re-used as reliable energy and supplies
Should.Therefore, the capability value of lithium ion battery can be chosen as the sensitive features of life prediction.
(3) capacity data pretreatment
On the premise of capacity of lithium ion battery degradation trend is not influenceed, smooth monotonic is carried out to its degradation in capacity data
Pretreatment, the degradation in capacity data after treatment are as shown in Fig. 4-7 orbicular spots.
(4) prior information of historical data is obtained
For lithium ion battery, the experience degradation model of its capacity is：
Qk=aexp (bk)+cexp (dk) (21)
In formula：A, b, c, d are model parameter；K is model variable, represents charge and discharge cycles number of times.
The curve that can be seen that degradation model from Fig. 4-7 curve-fitting results can well be fitted capacity and truly degenerate number
According to.Therefore, it can the empirical model by the use of its degradation in capacity as the measurement equation of prediction, i.e.,：
Qk=akexp(bk·k)+ckexp(dkK)+v, v~N (0, σv) (22)
Wherein, v is white Gaussian noise.
The state transition model and equation of lithium-ion battery systems are respectively：
xk=[ak bk ck dk]T (23)
In formula：wa,wb,wc,wdIt is white Gaussian noise, its standard deviation can determine according to model initial value.
Here by battery A1, the data of A2, A3 are used to obtain the prior information of information as known historical data.A4 conducts
Detection battery is used for life prediction.First by Matlab Curve Fitting Toolboxes to A1, A2, A3 data are fitted, obtain
Model parameter a, b, c, d (being shown in Table 1)；In order to reduce influence of the particular battery parameter to battery A4, a is chosen here, b's, c, d is flat
AverageAs the initial value of battery A4 degradation models.
The model parameter value of table 1
According to A4 battery model parameter a, b, c, d initial values, and noise is with respect to initial value
Can determine that plant noise is：wa~N (0,1e-10), wb~N (0,1e-6), wc~N (0,1e-5), wd~N (0,
1e-7), v~N (0,1e-3).
(5) prediction of monitoring object
It is N=1000 to take sampling population, and failure threshold is the 80% of battery rated capacity, is calculated according to particle filter is improved
Method is predicted to the monitoring objective (A4) of known portions capacity data.Fig. 8 is 18 longevity of measurement data before known battery A4
Life predicts the outcome, life prediction result when Fig. 9 is 30 capacity datas before known A4.The life-span predicted as can be seen from Figure 8
It is 48 circulations, predicated error is 1 cycle.PDF region representation predicting residual useful lifes probability density function (PDF,
Probability distribution function), its Breadth Maximum is narrower, and Surface prediction precision is higher, predicts the outcome more
Tool reference value.Final the predicting the outcome of Fig. 9 is 49 circulations, and error is 0, is illustrated with the known measurement data of monitoring object
More, residual life distribution is more concentrated, and it is also more accurate to predict the outcome.Detailed results are shown in Table 3.
Table 3 predicts the outcome detailed value
In order to illustrate embodiment compared to the advantage based on traditional standard particle filter life-span prediction method, here same
The two predicting the outcome under the same conditions is given on figure, as shown in Figure 10 and Figure 11, table 3 gives detailed forecasts knot
Really.As can be seen that under the same conditions, the precision of prediction of embodiment is substantially higher in the pre- of standard particle filtering from comparison diagram
Survey precision.Also, embodiment remained in the case where monitoring object given data is limited provide it is accurate predict the outcome, be therefore
The prevention of barrier provides reliable foundation.
It is pointed out that described above simply illustrate some principles of the invention, due to the general of constructed field
It is easy to carry out some modifications on this basis for logical technical staff and changes.Therefore, this specification be not intended to by
The present invention be confined to shown in and described concrete structure and the scope of application in, therefore every corresponding modification for being possible to be utilized with
And equivalent, belong to apllied the scope of the claims of the invention.
Claims (7)
1. the equipment method for predicting residual useful life of improved particle filter algorithm is based on, and the improved particle filter algorithm for being used includes
Importance sampling stage and resampling stage, it is characterised in that introduce Unscented kalman filtering method in the importance sampling stage
More new particle, generation suggestion distribution, so as to suppress sample degeneracy；Increase Markov chain Monte-Carlo step in the resampling stage
Suddenly, sample is suppressed exhausted.
2. the equipment method for predicting residual useful life of improved particle filter algorithm is based on as claimed in claim 1, it is characterised in that
The step of generation suggestion distribution, includes：
The initial variance of state value and particle according to previous moment particle calculates the sigma point sets of previous moment particle；
Bring the point that the sigma points of previous moment particle are concentrated into particle state equation, obtain the particle sigma points at current time
Collection；
Particle sigma point set weighted sums to current time, so as to obtain the expectation of current time particle；
The expectation of current time particle and the variance of the sigma point sets of current time particle are done, and gained variance is weighted
Summation, obtains the time updated value of the variance of current time particle；
The point that the point and the sigma points of previous moment particle that the sigma points of current time particle are concentrated are concentrated brings particle into
State equation, so as to obtain the observation of current time particle；
Observed value weighting summation to the particle at current time, obtains the updated value of the observation of current time particle；
Particle sigma point sets, the expectation of current time particle, the observation of the particle at current time based on current time, when
The updated value of the observation of preceding moment particle, is calculated the measurement updated value of the variance of the particle at current time；
Calculate the correction factor of the variance of current time particle；
It is the time updated value of the variance based on current time particle, the measurement updated value of the variance of the particle at current time, current
The correction factor of the variance of moment particle, is calculated the correction value of the variance of current time particle；
With the expectation of current time particle, the expectation and variance of the correction value as Gaussian Profile of the variance of current time particle,
Generation suggestion point.
3. the equipment method for predicting residual useful life of improved particle filter algorithm is based on as claimed in claim 1 or 2, and its feature exists
In Markov chain Monte-Carlo step includes：
Threshold value u~U [0,1] is extracted, sampling obtains the candidate state of particle from suggestion distributionUsing the candidate state of particleIts acceptance probability A is calculated, if u≤A, is receivedI.e.：Otherwise abandonCarry out resampling and reservation is adopted again
The particle of sampleI.e.：
4. the equipment method for predicting residual useful life of improved particle filter algorithm is based on as claimed in claim 3, it is characterised in that
Specific steps include：
(1) forecast model is set up；
(2) degraded performance is assessed, and chooses sensitive features as degenerative character；
(3) the original degenerative character chosen is pre-processed；
(4) on the basis of forecast model, prior information is obtained from the historical data of degenerative character；
(5) according to the initial information of each parameter in degradation model distribution and model, improved particle filter algorithm is initialized；
(6) estimation and two processes of outside forecast are iterated to each moment successively using improved particle filter algorithm, when moving back
When changing status predication value less than or equal to failure threshold, prediction terminates；
(7) state to all particle samples is counted, and obtains final predicting the outcome.
5. the equipment method for predicting residual useful life of improved particle filter algorithm is based on as claimed in claim 4, it is characterised in that
In the forecast model of step (1), predict that the system mode renewal equation of object is respectively with measurement equation：
xk=fk(xk-1,vk-1)
zk=hk(xk,nk)
Wherein, xkRepresent the system mode value at current time, zkExpression current time contains the measuring value of additive noise, vk-1For preceding
The process noise at one moment, nkIt is the measurement noise at current time, k ∈ N, N particle numbers, and process noise equal with measurement noise
It is stationary noise sequence.
6. the equipment method for predicting residual useful life of improved particle filter algorithm is based on as claimed in claim 4, it is characterised in that
Step (3) carries out the pretreatment of smooth and dullnessization to primitive character.
7. the equipment method for predicting residual useful life of improved particle filter algorithm is based on as claimed in claim 4, it is characterised in that
Prior information includes initial value of the degenerative character with each parameter in the distribution map and forecast model of variable.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710107434.5A CN106845866A (en) | 2017-02-27 | 2017-02-27 | Equipment method for predicting residual useful life based on improved particle filter algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710107434.5A CN106845866A (en) | 2017-02-27 | 2017-02-27 | Equipment method for predicting residual useful life based on improved particle filter algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106845866A true CN106845866A (en) | 2017-06-13 |
Family
ID=59133558
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710107434.5A Pending CN106845866A (en) | 2017-02-27 | 2017-02-27 | Equipment method for predicting residual useful life based on improved particle filter algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106845866A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107403279A (en) * | 2017-08-02 | 2017-11-28 | 中国石油大学(北京) | A kind of adaptive status early warning system and method for oil transportation pump condition |
| CN107403242A (en) * | 2017-08-10 | 2017-11-28 | 四川大学 | Based on the equipment method for predicting residual useful life for improving tasteless particle filter |
| CN110442941A (en) * | 2019-07-25 | 2019-11-12 | 桂林电子科技大学 | It is a kind of to be tracked and RUL prediction technique based on the battery status for improving particle filter and process noise features fusion algorithm |
| CN110456199A (en) * | 2019-08-14 | 2019-11-15 | 四川大学 | A kind of method for predicting residual useful life of multisensor syste |
| CN111044926A (en) * | 2019-12-16 | 2020-04-21 | 北京航天智造科技发展有限公司 | Method for predicting service life of proton exchange membrane fuel cell |
| CN111160666A (en) * | 2020-01-02 | 2020-05-15 | 西北工业大学 | Health state and reliability assessment method for monitoring strong noise and non-periodic state |
| CN111368403A (en) * | 2020-02-24 | 2020-07-03 | 西安交通大学 | Self-adaptive non-linear degradation residual life prediction method |
| CN111832125A (en) * | 2020-06-18 | 2020-10-27 | 东南大学 | Method for predicting fatigue crack propagation life of metal structure of metallurgical crane |
| CN112230154A (en) * | 2019-07-15 | 2021-01-15 | 中国科学院沈阳自动化研究所 | Lithium battery residual life prediction method |
| CN113504473A (en) * | 2021-09-10 | 2021-10-15 | 四川大学 | RUL prediction method suitable for power lithium battery nonlinear regression process |
| CN113657012A (en) * | 2021-07-21 | 2021-11-16 | 西安理工大学 | TCN and particle filter-based method for predicting residual life of key equipment |
| CN113987900A (en) * | 2021-10-26 | 2022-01-28 | 电子科技大学 | IGBT service life prediction method based on extended Kalman particle filter |
| CN114236414A (en) * | 2021-12-16 | 2022-03-25 | 上海交通大学 | Satellite lithium battery life prediction method based on KCC-PF algorithm |
| CN114357863A (en) * | 2021-12-13 | 2022-04-15 | 中国航发北京航科发动机控制系统科技有限公司 | Mechanical hydraulic product service life prediction method based on particle filtering |
| CN118013656A (en) * | 2024-02-02 | 2024-05-10 | 浙江大学 | Helical gear vibration response and rigidity degradation prediction method |
-
2017
- 2017-02-27 CN CN201710107434.5A patent/CN106845866A/en active Pending
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107403279A (en) * | 2017-08-02 | 2017-11-28 | 中国石油大学(北京) | A kind of adaptive status early warning system and method for oil transportation pump condition |
| CN107403279B (en) * | 2017-08-02 | 2020-05-08 | 中国石油大学(北京) | Oil transfer pump working condition self-adaptive state early warning system and method |
| CN107403242A (en) * | 2017-08-10 | 2017-11-28 | 四川大学 | Based on the equipment method for predicting residual useful life for improving tasteless particle filter |
| CN112230154A (en) * | 2019-07-15 | 2021-01-15 | 中国科学院沈阳自动化研究所 | Lithium battery residual life prediction method |
| CN110442941A (en) * | 2019-07-25 | 2019-11-12 | 桂林电子科技大学 | It is a kind of to be tracked and RUL prediction technique based on the battery status for improving particle filter and process noise features fusion algorithm |
| CN110442941B (en) * | 2019-07-25 | 2022-04-29 | 桂林电子科技大学 | Battery state and RUL prediction method based on particle filtering and process noise fusion |
| CN110456199A (en) * | 2019-08-14 | 2019-11-15 | 四川大学 | A kind of method for predicting residual useful life of multisensor syste |
| CN111044926A (en) * | 2019-12-16 | 2020-04-21 | 北京航天智造科技发展有限公司 | Method for predicting service life of proton exchange membrane fuel cell |
| CN111160666A (en) * | 2020-01-02 | 2020-05-15 | 西北工业大学 | Health state and reliability assessment method for monitoring strong noise and non-periodic state |
| CN111160666B (en) * | 2020-01-02 | 2023-06-23 | 西北工业大学 | Health state and reliability assessment method for monitoring strong noise and non-periodic state |
| CN111368403B (en) * | 2020-02-24 | 2022-03-08 | 西安交通大学 | Self-adaptive non-linear degradation residual life prediction method |
| CN111368403A (en) * | 2020-02-24 | 2020-07-03 | 西安交通大学 | Self-adaptive non-linear degradation residual life prediction method |
| CN111832125B (en) * | 2020-06-18 | 2021-03-19 | 东南大学 | Method for predicting fatigue crack propagation life of metal structure of metallurgical crane |
| CN111832125A (en) * | 2020-06-18 | 2020-10-27 | 东南大学 | Method for predicting fatigue crack propagation life of metal structure of metallurgical crane |
| CN113657012A (en) * | 2021-07-21 | 2021-11-16 | 西安理工大学 | TCN and particle filter-based method for predicting residual life of key equipment |
| CN113657012B (en) * | 2021-07-21 | 2024-07-30 | 北京智合瑞行能源科技有限公司 | Key equipment residual life prediction method based on TCN and particle filtering |
| CN113504473A (en) * | 2021-09-10 | 2021-10-15 | 四川大学 | RUL prediction method suitable for power lithium battery nonlinear regression process |
| CN113504473B (en) * | 2021-09-10 | 2021-12-10 | 四川大学 | RUL prediction method suitable for power lithium battery nonlinear regression process |
| CN113987900A (en) * | 2021-10-26 | 2022-01-28 | 电子科技大学 | IGBT service life prediction method based on extended Kalman particle filter |
| CN114357863A (en) * | 2021-12-13 | 2022-04-15 | 中国航发北京航科发动机控制系统科技有限公司 | Mechanical hydraulic product service life prediction method based on particle filtering |
| CN114236414A (en) * | 2021-12-16 | 2022-03-25 | 上海交通大学 | Satellite lithium battery life prediction method based on KCC-PF algorithm |
| CN118013656A (en) * | 2024-02-02 | 2024-05-10 | 浙江大学 | Helical gear vibration response and rigidity degradation prediction method |
| CN118013656B (en) * | 2024-02-02 | 2024-08-13 | 浙江大学 | Helical gear vibration response and rigidity degradation prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106845866A (en) | Equipment method for predicting residual useful life based on improved particle filter algorithm | |
| CN104778337B (en) | Method for predicting remaining service life of lithium battery based on FPCA (functional principal component analysis) and Bayesian updating | |
| Ma et al. | A new method for transformer fault prediction based on multifeature enhancement and refined long short-term memory | |
| CN103018673B (en) | Method for predicating life of aerospace Ni-Cd storage battery based on improved dynamic wavelet neural network | |
| CN108535656A (en) | Lithium ion battery remaining life prediction technique and system based on PCA-NARX neural networks | |
| CN107037373B (en) | Battery remaining capacity prediction technique neural network based | |
| CN112415414A (en) | Method for predicting remaining service life of lithium ion battery | |
| An et al. | A data-driven method for transient stability margin prediction based on security region | |
| CN114966233B (en) | Lightning forecasting system and method based on deep neural network | |
| CN113917334B (en) | Battery health state estimation method based on evolution LSTM self-encoder | |
| CN113205233B (en) | Lithium battery life prediction method based on wolf algorithm and multi-core support vector regression | |
| Shi et al. | Estimation of battery state-of-charge using ν-support vector regression algorithm | |
| CN112816874A (en) | RVM and PF algorithm fusion-based battery remaining service life prediction method | |
| CN114740388A (en) | Lithium battery residual life state evaluation method based on improved TCN | |
| CN113344288A (en) | Method and device for predicting water level of cascade hydropower station group and computer readable storage medium | |
| CN112287597B (en) | Lead-acid storage battery SOH estimation method based on VPGA-GPR algorithm | |
| CN108804334A (en) | A kind of discrete type software reliability growth test and evaluation method based on adaptive sampling | |
| Li et al. | A hybrid framework for predicting the remaining useful life of battery using Gaussian process regression | |
| CN114280490A (en) | Lithium ion battery state of charge estimation method and system | |
| Wei et al. | Remaining useful life estimation based on gamma process considered with measurement error | |
| CN115239971A (en) | GIS partial discharge type recognition model training method, recognition method and system | |
| CN104834816A (en) | Short-term wind speed prediction method | |
| WO2014173271A1 (en) | Optimization method and system for the number of monitoring units of digital man-machine interface | |
| CN116307139A (en) | Wind power ultra-short-term prediction method for optimizing and improving extreme learning machine | |
| Wang et al. | Remaining useful life prediction of capacitor based on genetic algorithm and particle filter |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170613 |